{"ID":2922220,"CreatedAt":"2026-06-02T02:42:49.606572591Z","UpdatedAt":"2026-06-02T23:57:59.442803696Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2606.00935","arxiv_id":"2606.00935","title":"Relational Intervention During Functional Collapse in Large Language Models: A Lexical-Statistical Ablation and a Structure x Register Factorial","abstract":"We test whether a relational-style intervention delivered during functional collapse in a small language model produces post-collapse behavior distinguishable from technical feedback, from a lexically-matched scrambled control, and from each of the two pragmatic dimensions in isolation. Using Qwen3.5-4B with a deliberately broken bash tool, we run 300 episodes across six conditions in a matched-pairs design (50 tasks): no intervention (A), technical/impersonal (B), relational/first-person (C), scrambled relational (D), technical/first-person (E), and relational/impersonal (F). E and F form a 2x2 factorial with B and C that dissociates relational structure (acknowledgment, absolution, agency restoration, unconditional acceptance) from sender register (first-person vs. impersonal). We report two main findings. First, an attention-behavior dissociation: attention follows lexical surprise (D \u003e F \u003e C \u003e E \u003e B, all q_FDR \u003c 10^{-10}), with the scrambled message capturing the most attention; yet behaviorally A ~ B ~ D \u003c E ~ F \u003c\u003c C. Second, the factorial localizes the C effect: neither relational structure alone (F) nor first-person register alone (E) replicates C's behavioral signature; main effects of both dimensions are individually significant, and the structure x register interaction is significant on persistence (p = 0.046). A third dissociation emerges in emotion probes: F tracks C on 7 of 8 probes despite producing only baseline behavior, indicating that relational structure alone installs a probe-level state that only translates into behavior when paired with first-person register. The model's processing decomposes into three dissociable stages: attention (ordered by lexical surprise), probe-level state (ordered by structure), and behavior (ordered by the conjunction of both).","short_abstract":"We test whether a relational-style intervention delivered during functional collapse in a small language model produces post-collapse behavior distinguishable from technical feedback, from a lexically-matched scrambled control, and from each of the two pragmatic dimensions in isolation. Using Qwen3.5-4B with a delibera...","url_abs":"https://arxiv.org/abs/2606.00935","url_pdf":"https://arxiv.org/pdf/2606.00935v1","authors":"[\"Franco Santana\",\"Horacio Vico\"]","published":"2026-05-31T00:10:01Z","proceeding":"cs.AI","tasks":"[\"cs.AI\",\"cs.CL\",\"cs.HC\"]","methods":"[\"Language Model\"]","has_code":false}